Problem. Conventional treatment does not restore normal motor function to many stroke survivors. The majority of available interventions direct treatment at the peripheral nervous system (arms/legs). Since stroke occurs in the brain and results in brain neuronal damage and dysfunction, a more direct approach would be to re-train the brain by directly treating the activation of brain signals that control movement. Purpose: Our purpose is to determine whether the surface-acquired brain signal (electroencephalography (EEG)) can feasibly be re-trained to drive more normal functional reach/grasp in stroke survivors. We will use two different and complimentary brain signal, training components. For the first component, we will train a progressively more normal brain signal (during upper limb reach components), in terms of four brain signal features: location of the signal, amplitude of the signal, wave form of the signal, and frequency content of the signal. In the second component, we will pair brain signal with the desired movement that is performed as close-to-normal as possible. Methods: Hypothesis I: Brain signal training will result in a more normal brain signal during a functional reach/grasp task. (Primary measure: EEG signal amplitude at electrode locations, C3/4 and C5/6, in the alpha frequency band (8-12Hz). Secondary measures will include: brain signal, event related desynchronization (ERD) at each of an array of electrode locations and at 1Hz frequency bins across 6- 30Hz. Hypothesis II: Specifically targeting, invoking, and training the surface-acquired EEG brain signal, and integrating brain signal training into motor learning training of the upper limb reach movement, will result in greater motor restoration versus a comparable comprehensive motor learning intervention without EEG brain signal training. (Primary measure: Arm Motor Ability Test, upper limb function). We will enroll 16 subjects who had a stroke (>6 months) and who will receive brain signal training and upper limb motor learning (8 with cortical;8 with sub-cortical stroke). We will enroll 8 additional control subjects receiving comparable upper limb motor learning, but no brain signal training. Treatment for both groups will be 5hrs/day, 5days/wk, for 12 wks, based on prior established motor learning protocols. For the 16 subjects in the brain training group, a single 5hr daily session, will be composed of: 1.0 hr/day, brain signal training;1.5 hrs/day, FES-assisted and robotics-assisted movement (no brain signal training included);2.5 hrs/day, motor learning (without modalities, without brain signal training). Significance: By directly re-training brain signal, the intervention has the potential to more completely restore motor function for more severely involved patients. PUBLIC HEALTH RELEVANCE: The purpose of this study is to determine whether non-invasively acquired brain signal (electroencephalography (EEG)) can feasibly be re-trained to drive more normal motor function in stroke survivors. Compared to conventional exercise, directly treating brain signal abnormality has the potential to be more beneficial for a greater number of stroke survivors and has the potential to more completely restore normal function, than is otherwise possible without direct brain training.